Case Study

How I Productized 15 Years of Marketing Experience With Claude as Infrastructure

Three tiers. Full automation. A real client tested end-to-end before launch. Here's what AI actually did, what I actually did, and why the difference matters more than most people writing about this admit.

By Shannon Ropelato, Founder, You Digital Co.

I spent the last few months turning fifteen years of marketing consulting into a productized service. Three tiers, priced at $197, $497, and $997. A Stripe checkout flow for each. A Make.com pipeline that takes a form submission, processes payment, generates a branded PDF audit, and delivers it without me touching anything. A KPI dashboard included with the executive tier, prebuilt and pre-populated. End-to-end tested with a real client (Smith Ashcraft, a Richmond family law firm) before any paying customer got near the system.

I built it with Claude.

That sentence is going to mean different things to different readers, so let me be specific about what it actually means before the takes pile up.

The work I directed

Five sessions in, Claude handed me a "KPI Dashboard Specification." It described tabs, columns, formulas, conditional formatting rules. It was thorough. It was technically correct. It was also commercially wrong.

A specification tells the client how to build a dashboard. A dashboard is a tool the client can actually use. Those are different deliverables. One is a document. One is a transformation. Anyone who has watched a consulting client's eyes glaze over while reading a spec they'll never implement understands the gap.

I sent one message back: "I am thinking this looks like something we should build the template out for and link to, it is a HUGE value add!"

That decision changed the executive tier from "expensive PDF" to "expensive PDF plus a pre-built tool worth $500 to $1,500 if you'd hired a freelancer to build it." Claude was thinking like a writer. I was thinking like a product strategist. The judgment was mine. The implementation, once I made the call, took Claude about thirty minutes.

That pattern repeated throughout the build.

When we designed the upgrade path between tiers, Claude initially proposed percentage discounts. I redirected toward fixed-dollar coupons keyed to VIP redemption codes. The economics are nearly identical. The psychology is not. A percentage discount cheapens the higher tier ("I'm getting 30% off"). A fixed-dollar VIP code feels like an insider perk ("you've been granted access to upgrade pricing"). One trains your buyer to wait for sales. The other trains them to feel chosen. Claude can read about that distinction in marketing books. Knowing when to apply it took fifteen years of watching premium positioning succeed and fail in real client work.

When Claude built an affiliate tracker, it produced a flat list of recommended tools. I rejected the structure. Recommending tools because they pay an affiliate fee, instead of because they're objectively right for the use case, undermines the consultative authority the audit is supposed to demonstrate. We rebuilt it as a tiered priority system. Tools get recommended on merit first, with affiliate links as a quiet secondary monetization layer that never compromises the recommendation itself. We don't lose trust to chase twenty dollars in affiliate revenue when the audit fee is $997.

There were smaller catches too. A competitor comparison table broke its own formatting because the firm name "Smith | Ashcraft" uses a pipe character, and pipes are markdown column separators. I caught it within two minutes of reviewing the draft. Not a strategic insight. Just the pattern recognition of someone who has built dozens of client-facing documents and knows how they should look. AI generates volume. Humans catch the things volume hides.

The most important decision I made was about timing. There were several points where I could have launched a partial product (just the starter tier, just the executive tier) and iterated in market. Claude would have happily helped me ship faster. I refused. Productized services live and die on first impressions, and I've watched too many consultants ship half-finished products they then spend a year apologizing for. We built all three tiers, the full automation pipeline, the dashboard, and the brand-aligned PDF generator. We tested all of it with a real client before opening the doors. That decision didn't come from the model. It came from scar tissue.

The work Claude compressed

Now the other side. Once I knew what to build and what "good" looked like, Claude compressed the work in ways I won't pretend I could match alone.

For the Smith Ashcraft executive audit, Claude researched seven competitor firms, pulled industry benchmarks for family law marketing (cost per lead in the $30 to $100 range, conversion rate medians, legal CPC ranges), and synthesized the findings into a forty-plus page audit running 13,878 words. Manually, that's twenty to thirty hours of research and writing. With me directing what mattered and Claude doing the volume, it took roughly two hours of active back-and-forth. Not because the model is magic. Because research-and-synthesize is exactly the kind of work AI does well when a senior person is calling the shots about what's worth synthesizing.

What "worth synthesizing" means in family law marketing isn't generic. Cost per lead matters more than cost per click for a firm that bills at retainer, because a $40 click that produces a $4,000 retainer client is a different math problem than a $40 click selling a $40 product. Average case value, lead-to-consult conversion, consult-to-retain conversion, and channel attribution by case type matter in a specific order, and that order changes by practice area. I know that order because I've worked the data for fifteen years across more than a hundred clients. Claude doesn't know which KPIs to weight heaviest until I tell it. Telling it requires the knowledge.

The PDF generator was a different kind of compression. I wanted my audits to render as branded documents (gradient cover pages in the YDC palette, automated headers and footers, page numbers, brand-aligned typography). That's not a marketing skill. It's a development skill, and a non-trivial one. The pipeline I now use runs Playwright for the cover page, falls back to wkhtmltopdf for content rendering, merges the result with pypdf, and uses CSS that survives both rendering engines. I could not have built that. I direct it through prompts. I maintain it through prompts. It exists because Claude was patient enough to debug a wkhtmltopdf gradient rendering issue at one in the morning when I'd given up.

The dashboard, once I'd decided it should be a real deliverable instead of a specification, came together in about thirty minutes. Six tabs, eighty-five formulas, conditional formatting with traffic-light indicators, brand colors throughout, sample data pre-populated so a new client can see how it works on day one. Zero formula errors on verification. A custom dashboard at that quality from a freelancer would run $500 to $1,500. I got it as a byproduct of the audit work.

The least visible but possibly most valuable compression was iteration speed. The audit master prompt (the document that tells Claude how to generate a tier-appropriate audit consistently) went through dozens of revisions across the project. Each revision was instant. I could test "what if the standard tier dropped this section" or "what if the executive tier opened with a competitive landscape instead of an executive summary" in thirty seconds, instead of rewriting documents. That's the part of AI that actually changes the economics of consulting. Not "AI does the work for you." Closer to "AI lets you test ten versions of the work in the time it took to produce one."

What the project taught me

The takes about AI replacing marketers are arguing about the wrong thing.

My value was never writing 13,878 words from scratch. A lot of marketers can do that. My value was never researching seven competitors manually. Interns can do that. My value was certainly never writing a Python PDF pipeline. Developers do that.

My value is the working knowledge base I've built over fifteen years and more than a hundred clients. I know which KPIs predict revenue and which ones flatter a dashboard without moving it. I know that for an ecommerce brand with a $60 average order value, ROAS is the conversation; for a family law firm at a $4,000 average case value, lead quality and consult-to-retain rate matter more than the front-end CPA most people fixate on. I know that for a piercing studio in a competitive metro, repeat visit rate and referral velocity tell you more about the health of the business than top-of-funnel traffic ever will. I know that benchmarks like "good email open rates are 20%" are useful only until you understand a specific client's list quality, and after that they're misleading.

A novice handed Claude and told to build the same audit could not produce what I produced. Not because the model would refuse to help them. Because they would not know which numbers to ask about, in which order, for which kind of business. They would accept Claude's first suggestion as authoritative because they wouldn't have the lived experience to push back. The audit would be thorough, well-formatted, and subtly off. The kind of off that costs a client thousands of dollars in misallocated budget while looking like a deliverable.

That's the part most "AI replaces marketers" arguments miss. The audit isn't the value. The framework that shapes the audit is the value. The framework is fifteen years of pattern matching across paid media, organic, email, conversion rate optimization, and analytics, knowing which patterns generalize and which patterns are specific to a vertical, and refusing to confuse the two.

If your value as a marketer is the output (the deck, the post, the deliverable), AI is a real threat. If your value is the working knowledge that decides what output is worth producing and what numbers are worth chasing, AI is a force multiplier you can't afford to ignore.

The marketers who panic about AI are usually the ones whose value is the work AI now does cheaply. The marketers who thrive are the ones whose value was always the layer above the work, the layer where you have to know things before you can ask the right questions.

That's not a comforting take if you've spent your career being judged by volume. It's also not a self-congratulatory one. Directing AI well is harder than it looks, and the discipline it requires comes from a place AI can't replicate: time. Hours logged. Clients won and lost. Campaigns that worked and campaigns that didn't, and the honest postmortems on why. There's no shortcut to that. AI doesn't shorten the runway to expertise. It rewards the people who already have it.

I caught myself more than once accepting Claude's suggestion because it was thorough and well-formatted, then realizing it was thorough and well-formatted in service of a deliverable that wasn't quite right. Direction is harder than approval. Direction requires knowing what right looks like before you see it.

What's actually live

For the readers who care more about proof than philosophy, here's what exists at the end of this:

This isn't a theoretical AI workflow. It's a product running in production.

Who this is for

If you're a marketing consultant or agency owner watching the AI conversation and wondering whether you should be worried, I'd offer this: the productivity story isn't the interesting one. The judgment story is.

The next twelve months will sort working marketers into two groups. The ones who use AI to produce more of what they already produce, faster. And the ones who use AI to take on work they couldn't take on before, because the things only they can do (strategy, judgment, taste, knowing what to build) are now the bottleneck instead of the typing.

I'd rather be in the second group. I built the YDC Audit Product to be in that group, on the operating side. The fight isn't human versus AI. The fight is judgment versus output. Pick the right side of that one.